nohz: Fix nohz ratelimit
Peter Zijlstra [Thu, 17 Jun 2010 16:02:37 +0000 (18:02 +0200)]
Chris Wedgwood reports that 39c0cbe (sched: Rate-limit nohz) causes a
serial console regression, unresponsiveness, and indeed it does. The
reason is that the nohz code is skipped even when the tick was already
stopped before the nohz_ratelimit(cpu) condition changed.

Move the nohz_ratelimit() check to the other conditions which prevent
long idle sleeps.

Reported-by: Chris Wedgwood <cw@f00f.org>
Tested-by: Brian Bloniarz <bmb@athenacr.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Greg KH <gregkh@suse.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Cc: Jef Driesen <jefdriesen@telenet.be>
LKML-Reference: <1276790557.27822.516.camel@twins>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

kernel/time/tick-sched.c

index 1d7b9bc..783fbad 100644 (file)
@@ -315,9 +315,6 @@ void tick_nohz_stop_sched_tick(int inidle)
                goto end;
        }
 
-       if (nohz_ratelimit(cpu))
-               goto end;
-
        ts->idle_calls++;
        /* Read jiffies and the time when jiffies were updated last */
        do {
@@ -328,7 +325,7 @@ void tick_nohz_stop_sched_tick(int inidle)
        } while (read_seqretry(&xtime_lock, seq));
 
        if (rcu_needs_cpu(cpu) || printk_needs_cpu(cpu) ||
-           arch_needs_cpu(cpu)) {
+           arch_needs_cpu(cpu) || nohz_ratelimit(cpu)) {
                next_jiffies = last_jiffies + 1;
                delta_jiffies = 1;
        } else {